Yousef Zayan

Key Skills

Software

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Text

Top Task Types

Evaluation Rating

Prompt Response Writing SFT

RLHF

Freelancer Overview

I am a highly motivated Computer Science and Artificial Intelligence student with hands-on experience in AI training data, specifically in mathematical reasoning and model evaluation. Through my work on projects like Kepler_RLHF Math, Advantage_Brain_MT Math, and Psychic – Kadabra with Outlier, I have contributed as both an Attempter and Reviewer, creating high-quality math prompts and solutions, evaluating model outputs, and ensuring logical accuracy across multiple reinforcement learning tasks. My expertise includes AI model fine-tuning, prompt engineering, and detailed task review in both Arabic and English. I’ve worked on diverse AI training projects involving reinforcement learning from human feedback (RLHF), multilingual content evaluation, and mathematical problem-solving tasks ranging from grade school to university level. My attention to detail, analytical mindset, and multilingual capabilities allow me to deliver accurate, clear, and valuable training data that helps improve state-of-the-art AI models.

Labeling Experience

Psychic – Alakazam (Math Prompt Generation and Review)

Scale AITextEvaluation RatingPrompt Response Writing SFT

Worked as a reviewer to assess original math prompts and solutions written in Arabic. Focused on clarity, correctness, and logical flow of tasks. Provided detailed justifications and feedback, ensuring data quality aligned with project standards for fine-tuning mathematical reasoning in large language models.

Worked as a reviewer to assess original math prompts and solutions written in Arabic. Focused on clarity, correctness, and logical flow of tasks. Provided detailed justifications and feedback, ensuring data quality aligned with project standards for fine-tuning mathematical reasoning in large language models.

2025 - 2025

Psychic – Kadabra (Math Prompt Generation and Review)

Scale AITextEvaluation RatingPrompt Response Writing SFT

Served as both Attempter and Reviewer in this SFT math-focused project. Created original math prompts in Arabic ranging from grade school to undergraduate difficulty. Wrote high-quality solutions and evaluated peer submissions based on clarity, correctness, and structure. Reviewed response quality and rewrote flawed answers to support high-quality model fine-tuning.

Served as both Attempter and Reviewer in this SFT math-focused project. Created original math prompts in Arabic ranging from grade school to undergraduate difficulty. Wrote high-quality solutions and evaluated peer submissions based on clarity, correctness, and structure. Reviewed response quality and rewrote flawed answers to support high-quality model fine-tuning.

2025 - 2025

Astrakhan_Grassland – Math Multimodal Evaluation

Scale AIImageRLHFEvaluation Rating

Evaluated the model’s ability to interpret mathematical expressions from images and word problems. Tasks involved recognizing mathematical symbols, parsing image-based equations, understanding natural language problems, and verifying the correctness and logic of AI-generated solutions. Contributed by creating structured prompts, testing edge cases, and rating the model’s accuracy and reasoning.

Evaluated the model’s ability to interpret mathematical expressions from images and word problems. Tasks involved recognizing mathematical symbols, parsing image-based equations, understanding natural language problems, and verifying the correctness and logic of AI-generated solutions. Contributed by creating structured prompts, testing edge cases, and rating the model’s accuracy and reasoning.

2025 - 2025

Advantage_Brain_MT Math – Multimodal Preference Ranking with Rewrites

Scale AITextRLHFEvaluation Rating

Served as a reviewer for preference ranking of AI-generated responses. Tasks included comparing multiple outputs from models in response to math/image-based questions, rating accuracy and coherence, and rewriting suboptimal responses to improve training data. Focused on model feedback in multiturn settings using mathematical and visual cues.

Served as a reviewer for preference ranking of AI-generated responses. Tasks included comparing multiple outputs from models in response to math/image-based questions, rating accuracy and coherence, and rewriting suboptimal responses to improve training data. Focused on model feedback in multiturn settings using mathematical and visual cues.

2024 - 2025

Kepler_RLHF Math – Model Failure Prompting and Rating

Scale AITextRLHFEvaluation Rating

Worked as an Attempter to create prompts specifically designed to reveal model weaknesses in mathematical reasoning. Evaluated AI responses, analyzed failure patterns, and rated the model's performance. Helped improve robustness and safety of LLMs by curating diverse failure cases and giving targeted feedback.

Worked as an Attempter to create prompts specifically designed to reveal model weaknesses in mathematical reasoning. Evaluated AI responses, analyzed failure patterns, and rated the model's performance. Helped improve robustness and safety of LLMs by curating diverse failure cases and giving targeted feedback.

2024 - 2024

Education

Work History